I need to feed input to an audio amplifier from an USB DAC. The audio amplifier has separated inputs for headphone and speaker but DAC has only one output (stereo). Moreover the speaker input is mono while the headphone input is stereo.

Since I don't know the input/output impedance levels of both devices I decided to use buffers with this arrangement: schematic

Is it correct?

Is an overkill?

There's a specific op-amp for such kind of job (TL081 seems good for audio management but I'm not an expert)?


The previous schematics was wrote in a hurry, this's almost complete... even if I have not decided whether to use buffers or not.

Updated schematics

P.S. Connecting DAC to power amplifier in this way still require a filtering stage with such large capacitors (100uF according to DAC application circuits) I tought to use such amplifier to avoid large caps on headphone output but I just ended up shifting them from side to side :-(.

---- UPDATE SCHEMATIC 17/11/2023 ---

Buffer-less schematic

I finally decided to remove buffers since I'm not skilled enough to manage the tuning of the resulting circuit. Hope trimming the last one will be easier.


Trying to verify if the passive network will preserve the signal I tought to use spice on an equivalent circuit and do an AC sweep between (20 Hz and 20kHz).

The simulated circuit is this:

Simulated schematics

The resulting amplitude graphs are these:

Result mag plots

Changing the input resistors on MAX9792 (putting all 4 resitors equals to 20k) will level all input to -3.6dB level I think it's correct since we are splitting signal on two equals parts after all.. correct?

Final input levels


4 Answers 4


TL081 seems good for audio management but I'm not an expert

TL07x and TL08x are really good and popular for audio applications. But before everything, your VCC is not given, and if it's below 5 VDC then you can't use TL08x as the minimum recommended supply voltage is 4.5V (for H suffix) or 5V (for others).

Is it correct?

Well, doesn't seem to be. Because, if you check the MAX9792's datasheet, you'll see that the HP amplifier section employs two common-offset inverting amplifiers with on-chip feedback resistors that are factory-trimmed to 40k2 ± 2k. So you need to feed the signals through a resistance (RIN2 = 40k2 for unity gain) in order to get proper operation/amplification. Similar thing applies to speaker amplifier section as well. Check the block diagram and, if given, typical application circuit schematic.

enter image description here

Is an overkill?

Buffering is always good but you may not need buffering unless the DAC's outputs are too high for audio amplifier IC (in that case you just divide the outputs, buffer them, and then feed to the amplifier through a resistor). Make sure there are coupling caps so the amplifier and DAC are DC-isolated from each other.

If you do need external buffering, make sure the output peaks stay below the TL08x's maximum swing which should be somewhere around ±0.5V (check the datasheet for correct values) because the DAC's output may have offset (VCOM is not connected in your schematic above so it's hard to guess).

If you check the block diagram above, you'll see that you actually don't need any external summer amplifier as the amplifier IC allows summing at the input. But of course, check the datasheet for further details and make sure the DAC's output levels are within the acceptable ranges of the amplifier.

  • \$\begingroup\$ I got most of your points ... but when you say IC allows summing at the input you refer to pin 1 an 28 where the differential inputs are shorted? \$\endgroup\$
    – weirdgyn
    Commented Nov 16, 2023 at 14:45
  • \$\begingroup\$ BTW Vcc = 5V ... DAC output level is between -0.3 and 4V (Vout) \$\endgroup\$
    – weirdgyn
    Commented Nov 16, 2023 at 14:50
  • \$\begingroup\$ @weirdgyn Sorry, I checked the correct block diagram before posting the answer but somehow I put the wrong one. Anyway. Now I updated the answer with correct block diagram (A and C suffix have the same config for SPK amplifier). As you can see, L and R channels can be summed at the input (pin28). That's how the IC allows you to sum the channels right at the input. So you may not need an external summer. \$\endgroup\$ Commented Nov 16, 2023 at 14:50
  • \$\begingroup\$ I tought it was a differential input management ... as it is for MAX9791 \$\endgroup\$
    – weirdgyn
    Commented Nov 16, 2023 at 14:57
  • \$\begingroup\$ @weirdgyn 9791 has stereo SPK amplifier but 9792 has mono. Even 9792 accepts summing at the inputs so it's regardless of the IC model. \$\endgroup\$ Commented Nov 16, 2023 at 15:04

U10 in the first schematic and U9 in the second schematic are not traditional inverting summing circuits. They are non-inverting summers. This circuit does *not present a virtual GND at the summing node. Because of this, if the driving point impedance for the summing resistors is not zero ohms, the input signals can cross-contaminate each other. Since you state that you don't know the output impedance of the DAC, this could be a problem depending on how demanding you are about the "purity" (or whatever) of the audio.

A solution is to move the two summing resistors (R36-37 or R47-48) to the outputs of the left and right channel signal buffers. These are clean, low impedance driving points with output impedances way lower than the 10K summing resistors.

But Wait . . .

Based on the block diagram, it looks like the MAX9792 already has a true, inverting summing amplifier built into the SPKR input signal path. Note that the left and right signal paths also have inverting amplifiers in them, preserving signal phase between the two sets of inputs. Did I miss something in a posting, or does this eliminate the need for an external summer?

There are two Cin2's in the Maxim image, probably a documentation error. These would be C44 and C45 in your second schematic.

Connect left side of C46 to the output of U10, and add another R and C from the 9792 pin 28 to the output of U11. I think this is the intent of the Maxim part. R52 and the new resistor do not have to equal R50 and R51. You can use these resistor values to adjust / scale / balance the stereo/mono signal levels out of the 9792.


There are different configs possible, You can mix at the input of 3rd opamp or use a 4th and mix at their outputs.

There are always trade offs, for high speed its better to load it more but this will worse other things.

Without getting in details I see you get low distortion for RL >10k for the dac, so if design is mine, I will load it to 24k (little higher)

With your configuration you can achieve it with one 47k on positive stereo op amp inputs And mix section impedance of 47k which first op amps sees as ~24k To make it balanced 3rd one must "see" same impedance so you must put 24k resistor to ground there with capacitor in series to not divide the bias voltage. At this point imagine you play mono signal on dac The mix resistors are like getting from same source in parallel. So to achieve the needed 47k mix section These must be 47k-24k as once, this is 24k when paralleled or 47k as single resistors. The problem with this is you get half signal and you must use *2 gain also with capacitor in feedback or dont use these two capacitors i mentioned and divide/amplify bias too.

As you see this way you get much and complex choices and go way too deep in details which distract you from your main focus. So you must use this:

(simple) Or seems better to use 4 op amps with jfet inputs and just use 22-33k on both audio rails and connect positives together and mix signals at the outputs of 3rd and 4rd opamps through 150 ohm resistors to get the standart 75 ohm impedance for RCA Or higher if you dont want to load the amp but remember about the trade offs.

Its about specific choice and own taste, stragedy.

Rohat told you about the other inputs.

PP: I calculated Rins for your configuration and draw a schem for you. Somebody said to AC couple, i not see reason for that, i prefer to use the dac to bias the op amps and save resistors.

You have 5V supply, dac outputs 55% Vcc Lets say 1.8 gain HP and SPK are different because internal input difference in design of IC


simulate this circuit – Schematic created using CircuitLab

  • \$\begingroup\$ Please use the circuit tool to draw a diagram, edit the question \$\endgroup\$
    – Voltage Spike
    Commented Nov 16, 2023 at 22:53
  • \$\begingroup\$ thnx I' trying to understand this circuit... \$\endgroup\$
    – weirdgyn
    Commented Nov 17, 2023 at 11:21
  • \$\begingroup\$ its same but mixed at the output, maybe crossed connection confuses you, \$\endgroup\$
    – acho
    Commented Nov 17, 2023 at 14:22

The buffers for the other paths are maybe not necessary but not a bad idea, however see my comment below on power supplies. If you use a quad op amp (eg TL074) you have the stages available anyway. However I would generally use inverting amps because you can change resistor values later to match signal levels between devices, which might be necessary. However ...

Power supplies look to be an issue. What voltage is Vcc? take care about biassing and levels of power supplies. This doesn't look so good. TL0xx op amps typically need balanced supplies, and then you are going to have the issue that the bias points between stages are incorrect. If you are trying to run everything off one supply, you might be better looking carefully at the datasheets for the devices and seeing if you can directly connect them. The L+R mix might be possible using a simple resistive mixer.

  • \$\begingroup\$ you're absolutely right about U10 feedback resistor... it was a typo of course.... Vcc = 5V anyhow but I got the point ... \$\endgroup\$
    – weirdgyn
    Commented Nov 16, 2023 at 14:39
  • \$\begingroup\$ ok so if you are going to do it with opamps, you need something that will work on +/-2.5V. I did look at the datasheet for the DAC, the output voltage is rather low and centred on 0.55v I think. I did not find the spec for the power amp, you need to find that and design accordingly. You cannot skip this step. \$\endgroup\$
    – danmcb
    Commented Nov 16, 2023 at 14:56
  • \$\begingroup\$ I'm still trying to figure it out... I don't want a complex circuital solution because it's a pretty small device and otherwise it has not to be HiFi ... \$\endgroup\$
    – weirdgyn
    Commented Nov 16, 2023 at 15:00
  • 1
    \$\begingroup\$ "First off, U10 needs a 10k feedback resistor. As it stands there will be no output." I don't see that. First off, U9, U10, U11, and U12 do different things in the two schematics. Please fix that. Second, first sch U10 and second sch U9 are not traditional inverting summers. They are non-inverting summers, and will work just fine with a voltage follower. They have another issue which is in my answer. \$\endgroup\$
    – AnalogKid
    Commented Nov 16, 2023 at 15:04
  • \$\begingroup\$ ok I fixed the second schematics right now... \$\endgroup\$
    – weirdgyn
    Commented Nov 16, 2023 at 15:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.